Radio Astronomy SA

From the outset, a reliable solar power system was going to either make or break the project.  There is mains power running across the property in the form of a Single Wire Earth Return (SWER) line, courtesy of SA Power Networks but it was far enough away for us to conveniently ignore, for a host of reasons.  In our solar power system, the power buss would be pretty simple, everything would operate from a nominal +12VDC system.  It is important to remember that a 'nominal +12VDC system' really means that, depending upon the level of charge in the battery,  the nominal +12VDC can actually be anywhere between +11.8VDC and +14.4VDC at any time.  Variations in voltage translate into changes in power consumed/dissipated for a given current consumption of the equipment that we use.  A couple of hours after mid-day on a random day recently, we have records of the battery voltages being +14.0V on all three solar systems (see below).  This data will be used to calculate power consumption for 'real world' examples on this webpage.

A pair of 80W PV panels were installed above the SpaceCase on the tower and within the SpaceCase, a Victron Energy MPPT 75|10 Solar Controller was chosen to charge the 90AH lead-acid cranker (automotive) battery located under the tower at ground level.

For Phase 1, the commercial-off-the-shelf (COTS) Callisto Receiver draws ≈260mA @ +14VDC in measurement mode, power consumed is 3.7W.  The Callisto-Lx computer draws ≈130mA @ +14VDC and does include DC-DC converter losses, power consumed is 1.9W.  The DC-DC converter output is fixed for a +5VDC input to the computer.  The 2x LNAs each draw ≈90mA @ +14VDC, power consumed is 2.6W combined.  The 3G Internet Access Point draws ≈200mA @ +14VDC, power consumed is 2.8W.  The estimates shown below have been rounded up by the online power calculator to present a worst case scenario. 

Both instruments and ancillaries ran 24/7, so total power consumption was ≈264 Watt Hours/day.   For 3 days of autonomy we would need a 155Ah lead-acid battery to maintain operation without disruption.  Assuming that we can get 6 sun hours of charging per day, we would need at least 165W of PV panels to charge the battery.  Continuous operation with 6 sun hours per day through our winter months would be optimistic while the LPDA antenna was parked in transit configuration.  Once the solar tracker was turned on, things became less reliable and we lost a day or two occasionally in June/July - our darkest months.  Fortunately, we were still ironing out issues with other things on-site so we didn't need rescue missions, site visits were still a regular thing.

There had been a software provision to switch the COTS receiver, LNAs and Internet AP off overnight using a solid state relay but it was never implemented. 

We began by looking out for domestic PV panel installation upgrades, a cheap source of larger, used panels that we could buy for around $40 each.  The 75V Victron MPPT controller was upgraded from 10A to a 15A output model, to allow a pair of 60 cell, ≈200W PV panels to be configured in either series or parallel.  Series connection provides higher PV output voltages which generally maximises power conversion under off axis operating conditions (early morning, late afternoon) for fixed panels facing due North.  We had previously been using a cranker battery out at the tower and this was replaced with a 100AH deep cycle lead-acid from ALDI.  The two 80W PV panels were replaced with a single 200W panel, located above the SpaceCase.  By now, David & Tique's container installation had been completed, their solar power system had been installed with 2 x 160W panels on the roof and 2 x 90A deep cycle batteries with a 30A MPPT controller inside.  In preparation for the introduction of the Phase 2 & 2.5 spectrometers, the Phase 1 Callisto-Lx computer was moved into the container, we had already swapped out the 3G Internet AP for the data link back to the house.  This decreased the daily power consumption out on the tower considerably, bringing the load on the battery back to a more sustainable level.  This was to be short lived because Phase 2.5 was introduced soon after.

Phase 2.5 would bring a considerable increase in power consumption and we will need to spread the load across different locations/DC power sources.  Out on the tower, the 200W PV panel and original frame were removed and a pair of 190W panels were mounted on a heavy frame located on the northern side of the tower.  The Phase 1 Callisto-Lx computer was removed and an industrial mini-ITX Windows PC was installed in the container to control both Phase 1 & Phase 2 Spectrometers.  The new 2x COTS receivers each draw an additional 260mA @ +14VDC in measurement mode, power consumed is 7.4W combined.  The new 2x COTS FEEs consume 230mA @ +14VDC each, power consumed is 6.5W combined. and the Local Oscillator (LO) in the Up-converter draws 300mA @ +14V, power consumed is 4.2W.

With the introduction of another two receivers, we had a few issues with the new industrial computer not being able to handle all the data from the 3 COTS receivers.  We were seeing the occasional corrupted file and that prompted some serial communication changes back to the container, which added another box into the SpaceCase.  At the same time, we added Blair's autonomous GPS switch, which would be used to power down all three receivers, the two LNAs, the two FEEs and the LO overnight.  Both the serial converter and GPS switch added some additional power drain but it was a vast improvement on leaving everything powered up overnight.  If we had ran all Phase 1 & 2.5 instruments and ancillaries 24/7, the power consumption would be ≈624 Watt Hours/day.  With the introduction of the GPS switch, this reduced to ≈450Wh/day, the daytime consumption being ≈420Wh and the overnight power consumption reduced considerably to ≈32Wh.

We also chose to add another pair of 190W PV panels, a Victron MPPT 75|15 Solar Controller and a 90AH cranker battery to the container to help out with the load, and in anticipation of Phase 3.

Now there were three solar systems in operation that needed to be managed, ideally on a daily basis and just as ideally, from a considerable distance away,  like from home.  Fortunately, Blair likes to build stuff so he set off to build a multichannel data logger that could be used to monitor the battery state of all three solar systems.  The first industrial computer had failed and not too long after replacing it, so did the second industrial computer.  Having become accustomed to logging into the site with TeamViewer and then suddenly losing that capability,  something would need to be done to rectify these problems.  Peter had been trialing a debian version of Callisto_for_Unix on a Raspberry Pi 4B, primarily because our old Pi Model B+(s) were becoming difficult to support.  The Pi 4B with AnyDesk installed was quickly deployed as a remote gateway into the site and it also replaced the Pi Model B+ that we were using to control the Phase 2.5, Focuscode 63 Spectrometer.

For Phase 3, the 2 x LVM Callisto Receivers each draw ≈165mA  in measurement mode with the DC-DC converter output set for a +7.5VDC input to the receivers.  Power consumed is 2.5W combined but does not include DC-DC converter losses from +14VDC.  The 2 x Callisto-Lx computers each draw ≈130mA @ +14VDC, power consumed is 3.8W combined and does include DC-DC converter losses.  The DC-DC converter output is fixed for a +5VDC input to each computer.  The 2 x LNA pairs each draw ≈150mA with the DC-DC converter output set for a +6.0VDC input to the LNAs.  Power consumed is 1.8W combined but does not include DC-DC converter losses from +14VDC.  The estimates shown below have been rounded up by the online power calculator to present a worst case scenario. 

Both instruments and ancillaries ran 24/7, so total power consumption was ≈192Wh/day.  For 3 days of autonomy we would need a 113Ah lead-acid battery to maintain operation without disruption.  Assuming that we can get 6 sun hours of charging per day, we would need at least 121W of PV panels to charge the battery.

We have had some issues over the past couple of years with some data getting slightly messed up, and had some suspicions that it might be related to batteries / power supplies.  Blair installed the voltage data logger made up of an ADM 4280 -C data acquisition module that communicates via Modbus (a form of RS422).  These modules are about $40 on Alibaba and given the success that he achieved with a data acquisition system (DAQ) installed at our TPSO Seismic vault to monitor temperature and humidity and to control the humidity there, it seemed that this addition to Sunnydale might allow us to monitor the batteries, the performance of the 3 separate solar power systems and maybe track down any power related issues we might have.  All 3 solar / battery systems have the same solar panels, lead-acid gell cells, however the charging systems are different.  The one at the tower and our system in the container have Victron MPPT 75|15 chargers and these we are pretty happy with.  The 3rd system in the container had a Powertech MP3741 MPPT Solar Charge Controller installed but while it seems to work, when we upgraded the solar panels a couple of months ago, we started to see more data sync issues.

 The installation of the ADM module was in part to try and track down the issue and a good thing to have to actually monitor the battery systems anyway.  Blair wrote some python code to talk 'Modbus' to the ADM module runing on a Raspberry Pi 4B and modified the ADM module from current mode to voltage mode.  Some bench testing with a Fluke voltage calibrator was done to calibrate each channel, the voltage resolution is 1mV and probably accurate to ±10mV, depending on the ambient temperature in the container.  Perfect (overkill?) for our needs.  Initially, the data acquisition module ran at 1 sample every minute, and managed to capture a 16.8 volt spike on the output of the Powertech unit.

 Progress!  It seems that the Powertech unit might not be ideal for our usage after all and perhaps the reason that two industrial PCs died in use within the container.  The python code was changed to run at 1 sample per second for a few hours each morning (it could run that way 24 hours a day), but it generates a few Mb each day and storage might become an issue.  We captured several voltage spikes during the next month and they seem to be related to when the Powertech unit was in 'bulk charge' phase (early in the morning) and when there were clouds passing the site.  We suspect the averaging loop for the MPPT cycle in boost mode is done over a too long a period and when the Sun pops out from behind a cloud, the charger takes too long to respond and hence, we get the voltage spikes, they only last for a few seconds, but that's enough to mess with some of the data.  So, armed with data, we decided to replace the Powertech unit with another Victron MPPT 75|15.

Now we can sit at home, log into the site and see what is happening in real time and be able to graph it over long periods using TSoft, a software package normally used in our some of our seismic activities.

This 24 hour plot of all three battery voltages is generated in TSoft from the three active channels from Blair's logger.  This plot begins at 00:00UT, the three Victron 75|15 controllers are close to completing their 'bulk charge' mode. 'Absorption charge' mode commences and lasts until 08:00UT (dusk) and then they change to 'float mode'.  Float ends around 21:30UT (daybreak) and 'bulk charge' begins to recharge the batteries again.

Keen observers will have noted the voltage bump in the Tower (V) plot beginning at 10:30UT and finishing at 18:30UT.  This characteristic is the result of the autonomous GPS Switch shutting down the two Callisto Receivers, two LNAs & Up-Converter Local Oscillator located in the Spacecase mounted on the Tower.  This saves a considerable amount of power from being wasted overnight.